1. Field of the Invention
The present invention generally relates to data processing for searching of databases and, more particularly, to graphic interfaces for facilitating evaluation of and navigation through groups of items such as results of searches containing potentially related information not necessarily responsive to a search query such as documents, articles and the like which may cite other such documents, articles and the like and which has particular utility for legal, medical, scientific and similar types of research.
2. Description of the Prior Art
Since the development of the digital computer, a principal use for data processing systems has been for managing and retrieving data of interest from one or more databases including potentially vast amounts of information. Expediting the return of search results has been a major goal of such systems and many approaches to the development of efficient search engines and algorithms have been proposed with greater or lesser degrees of success, which often varies with the nature of the subject matter of the data in a given database as well as the size of the databases which are searched and the processing power available. In general, such search engines must also be able to accommodate a search query that may be in a fairly rigidly expressed form (e.g. key words with Boolean operators) but yet can express an arbitrary search query and also provide for manipulation of the search query (e.g. logical combinations of search queries in accordance with Boolean operators) to provide some degree of control to the user over the unavoidable trade-off between the number of returned search results, the relevance of the information returned and the likelihood that highly germane information will not be returned in response to a particular search query.
Unfortunately, to be reasonably certain that all highly germane information is included in the search results, the search query will usually be required to be of sufficient breadth that the search results themselves may be quite numerous. The searcher will then have no alternative but to review the entirety of the search results to (often subjectively) determine the most relevant information which is actually sought among the search results with little or no guidance as to the relative likelihood that a given article or document will be the document or article actually sought or even which group of articles or documents will be most likely to contain it. This lack of guidance is largely because of the trade-off noted above which implies a high likelihood of exclusion of the article or document actually sought by any limitation of the search query or results.
To expedite this process while reducing the likelihood of exclusion of germane results, numerous sorting algorithms and techniques have been proposed and which have also provided widely varying degrees of success. Most such sorting arrangements, in practice, merely estimate a rating or a rank of the individual search results in accordance with a degree of accuracy of match to the search query that produced them and are thus highly dependent on user input based on relatively minimal information in regard to the actual content of the database; resulting in only marginal effectiveness in many cases. Often such sorting algorithms and methodologies are based on citations of other documents or articles in the database and seek to determine relevance from some combination of or relationship between the documents or articles containing citations to other particular documents or articles and/or the number of citations to particular documents or articles between different groups of documents or articles such as between databases or results of different searches or the like. However, one particularly effective sorting arrangement is disclosed in U.S. patent application Ser. No. 09/707,911, entitled “Improved Relevance Sorting for Database Searches”, filed Nov. 8, 2000, by Edward J. Walters et al. and which is hereby incorporated in its entirety by reference. While the sorting provided by this arrangement allows very substantial control over sorting and re-sorting of the search results while allowing the user to compare results of a sort to determine further searching and/or sorting options, the result remains in the form of a list or lists which may not be optimally assimilated by a user. For example, as a practical matter, a displayed list can only show a limited number of responses to a query that may be a very small proportion of the results returned. The efficiency of the sorting arrangement may thus cause highly relevant material to be distributed over several screens; thus necessarily excluding some relevant items from any given screen and removing them from view while the user attempts to refine the search or sorting of results if such facilities are provided. Further, some information that could be important in refining the search or sorting of results may not be directly responsive to a search and cannot be easily portrayed in a manner easily assimilable by a user.
In this regard, it has long been recognized in the data processing arts that the amount of processing power and the efficiency of algorithms that can be brought to bear on a particular data processing task are of little effectiveness beyond the ability or capacity of a user to assimilate the results of the processing performed. This fact is particularly characteristic of database searches in legal, medical and scientific fields where there are both objective and subjective aspects of determination of the most germane information within particular search results. For example, while powerful computers running the most efficient and effective searching and sorting algorithms, determination of the most germane information may still require critically reading through and comparing at least a significant number of the returned documents or articles (or, in the field of legal research, published decisions which are intended to be included within the terms “articles”, “documents” or “items”) rated or ranked as most highly relevant (e.g. by estimating degree of match to the search query by the numbers of instances of appearances of search terms or associations of search terms, combinations of citations or the like criteria) to determine those which are the most authoritative as well as the relationships to subject matter and the relationship of decisions rendered to other reported decisions. Such a detailed review may thus require many times longer than the searching and sorting processes and, to date, no system or interface has proven significantly more effective than a sorted list which requires such review and which is subject to error (or differences in estimated match to the search query as an indicator of likely relevance based on insignificant differences in scores made for ranking search results) while not supporting any application of subjectivity or expertise on the part of the searcher other than through such a detailed and critical review.
Additionally, in many fields, and legal research in particular, several different types of criteria may be applied to determination of relevance, authority and other qualities (e.g. an indication of the history and/or development of the line of reasoning represented) which may be of interest in the information sought or may indicate that a particular document or group of documents is germane to the search, whether or not included in a particular set of search results, and thus may require multiple dimensions (both qualitative and quantitative) to represent the information which may be available and of interest. Such information would necessarily be omitted or obscured in known displays of lists which only show one screen at a time; each screen containing on a partial list of search results.
Unfortunately, while visual displays have proven, over many years, to be the most effective as well as most efficient medium for an interface to present information to a user, such displays are limited to two (e.g. Cartesian) dimensions. While numerous graphic features (e.g. color, textures and the like visual effects, sometimes collectively referred to as attributes hereinafter) are known as well as various juxtapositions of data for representing more than two dimensions simultaneously, even the most skillful interface design cannot guarantee that the information can be assimilated rapidly, accurately or reliably by a user in order to be useful in selecting particular data of interest from among the potentially large amount of information that may be returned in response to a query and presented to a user or which is to be otherwise analyzed using the interface. In other words, while attributes have been used to convey additional information beyond two dimensions they have done so in a largely non-intuitive manner which does not aid in assimilation of information and, moreover, as provision is made for portraying more types of information, do so at the substantial risk of developing image complexity which can obscure information. Also, even if data is not obscured by image complexity, while a very rich amount of data and interrelationships may, in theory be presented, the user is only provided with relatively minimal control over the display and, hence, the level of expertise which must be brought to bear on the analysis process is extremely high. To date, no such multi-dimensional display has proven effective for recognition of particular relevance in combination with authority and other qualities of documents in a search result and which also provides highly intuitive control and selective data suppression to facilitate human analysis, particularly for the purpose of refining ordering of search results for detailed review and expediting discovery of highly germane documents included in the search results.